All workCase 04 / 06

DeFAI

SAM Terminal

LivePM · 2025–2026samterminal.com ↗
TypeScriptGoNestJSTurborepoDockerPostgreSQLRedisMCP
Plugin runtime · MCP · multi-chain

Open-source DeFAI framework for on-chain trading agents.

Users & job-to-be-done

Builders and agent operators who need a reusable trading runtime — CLI, MCP and OpenClaw skills — not another locked consumer app.

Problem

Teams rebuilt wallets, execution, scheduling and tool wiring for every trading-agent idea. Hosted consumer apps also trapped value inside a UI most builders outgrew.

Insight

The durable product was infrastructure: plugins, DAG workflows, MCP tools and skill files. Consumer UI was a distribution detour, not the end state.

Bet

Pivot from hosted consumer to a self-hostable OSS framework with MCP-native tools, multi-chain execution and first-class builder surfaces (CLI, MCP, OpenClaw).

My role

PM ownership of vision, strategy, roadmap and PRDs. Led the consumer → OSS pivot with engineering (TypeScript/Go monorepo), and shaped docs/GTM so builders could self-host without a white-glove onboarding team. High autonomy in a fast TinyLabs environment.

Journey

  1. Setup: clone, env, Docker/db, pnpm build
  2. Compose: enable plugins (swap, token, wallet, AI, Telegram…)
  3. Connect: CLI, MCP server, or OpenClaw skill
  4. Run: schedule / execute DAG workflows with risk controls

Architecture

  • MCP server exposing 40 tools (tokens, wallets, swaps, flows) to AI agents, backed by gRPC services.
  • Plugin contract (actions, providers, evaluators, hooks) driving a DAG flow engine with cycle validation.
  • Turborepo monorepo: TypeScript core plus NestJS and Go services on shared protobuf contracts.

Tradeoffs

  • MCP-native tools and plugins over a locked consumer UI.
  • CLI, MCP and OpenClaw as equal builder surfaces.
  • Sunset hosted consumer and reopen as self-hostable OSS.

Validation

  • Hosted-product pain: builders kept asking for programmable guts, not more consumer chrome.
  • Packaging tests: docs, skill files, CLI and MCP paths as the real “activation” funnel for infra.
  • Architecture review against production concerns: gRPC services, dry-run, graceful shutdown.

Outcome

North star: Time-to-first working agent for a builder on the OSS path

40+
MCP tools
6
plugins
6
EVM chains
DAG
Workflows
OpenClaw
skills

A self-hostable DeFAI framework — plugin runtime, DAG workflows, MCP tools and OpenClaw skills — so builders ship trading agents without reinventing the stack.

Learning

Consumer traction without infrastructure depth is a local maximum. The honest PM call was to sunset the hosted surface, preserve trust on the way out, and reinvest in the runtime builders actually needed. That pivot is the product.

Skills & know-how

Product strategy
Vision, roadmap, hard pivot.
PRD writing
Specs eng could ship from.
Prioritization
Consumer chrome vs runtime depth.
X-fn delivery
Eng, docs, OSS GTM.
AI-native infra
MCP + OpenClaw skills.
Dev-tool PM
Built for builders, not tourists.

What this proves: that I can own product strategy end to end — including the uncomfortable pivot from consumer app to developer infrastructure — and still ship a coherent MCP-native, multi-chain framework.

Open to work · Full-time, contract, advisory

Hiring a product engineer who ships?
Let's talk.

Product, engineering or go-to-market roles, remote from Türkiye (UTC+3). Send the job description or grab 15 minutes.

© 2026 Mesut Gülecen